Cows select between plants during grazing and such selecting behavior is also observed when feeding total mixed rations (TMR). Cows prefer smaller particle size (grains) to fibers (Miller-Cushon and DeVries, Reference Miller-Cushon and DeVries2017). Feed sorting can lead to imbalanced energy input and lower nutritive value of leftover TMR (DeVries et al., Reference DeVries, von Keyserlingk and Beauchemin2005). Consumption of suboptimal amounts of fiber and higher amounts of readily fermentable carbohydrates can result in subacute rumen acidosis (Plaizier et al., Reference Plaizier, Krause, Gozho and McBride2009), while an energy deficient ration induces subclinical ketosis (Oetzel, Reference Oetzel2007). Both forms of metabolic disorders are highly prevalent in many European dairies (Suthar et al., Reference Suthar, Canelas-Raposo, Deniz and Heuwieser2013).
TMR sorting can be detected by fraction measurements using a 3-sieve Penn State Particle Separator (Maulfair and Heinrichs, Reference Maulfair and Heinrichs2010). TMR separation allows to estimate the forage to concentrate ratio before and after feeding, the chop length of each fiber fraction and the peNDF content on the basis of NDF content of the TMR.
Our hypothesis was that the presence of feed sorting is associated with a higher incidence of imbalanced energy intake manifested in higher prevalence of subclinical ketosis and acid load. We have tested our hypothesis by comparing main metabolic parameters and the prevalence of metabolic disorders between herds where feed sorting was present and absent, respectively.
Materials and methods
Farms, animals, sampling
Within the course of regular herd health and metabolic profile monitoring, 22 Hungarian large-scale dairy herds were visited. Blood and urine samples were taken from 526 clinically healthy cows from the following production groups: fresh cows (1–7 d postpartum (pp), n = 74), early lactating cows (8–30 d pp, n = 230) and peak lactating cows (60–100 d pp, n = 222). Samplings took place 3–5 h after the morning feeding. Blood (25 ml) was collected from the udder vein (v. epigastrica cranialis) into heparinized tubes, urine samples (25 ml) were collected by means of a catheter from the urinary bladder. Samples were immediately cooled to 4 °C and transported to the laboratory, where the blood samples were centrifuged (4000 rpm for 10 min) and then analyzed. Plasma beta-hydroxybutyrate (BHB) was measured by an A-25 analyzer (Biosystems S.A., 08030 Barcelona, Spain) with a colorimetric assay kit (DCC430100, Diagon Ltd., 1047 Budapest, Hungary). Net acid base excretion (NABE) of urine was assessed according to the method of Kutas (Reference Kutas1965). Subclinical ketosis was defined as BHB concentrations >0.8 mmol/l in blood plasma, and acid load (subclinical acidosis) was defined as NABE <100 mmol/l in urine.
TMR separation
Feed sorting was determined on the basis of differences in the mass proportion of TMR fractions between freshly distributed and leftover TMR. At the time of morning distribution of TMR (between 5.00–6.00 h), an approximately 1–1.5 kg amount of TMR was collected from the feed trough of each production group (by mixing handful amounts collected from 4–6 sites along the trough) at each visited farm. TMR samples were fractionated by a 3-sieve separator (Penn State Particle Separator; Heinrichs, Reference Heinrichs2013). According to Heinrichs (Reference Heinrichs2013), recommendations of mass proportions of fractions 2–8% should appear on the upper sieve (ø > 19 mm), 30–50% on the middle sieve (ø 8–19 mm), 10–20% on the lower sieve (ø 4–8 mm) and 30–40% in the bottom pan (ø < 4 mm).
The same sampling and fractionating procedures were repeated with orts (at 12.00–13.00 h, 6–8 h after the feed distribution and immediately after blood sampling).
If the mass proportion of any of the TMR fractions showed a difference of at least 5% between fresh and orts, it was evaluated as an evidence of feed sorting (Heinrichs, Reference Heinrichs2013).
Statistical analysis
Cows were considered as ‘(having opportunity for) feed sorting’ if they belonged to a group where TMR separation indicated feed sorting, and ‘non-sorting’, if the TMR separation indicated no feed sorting. A dataset was compiled that included the biochemical parameters, ketosis/acid load status, production group, herd and ‘feed sorting status’ of each cow that was sampled during the farm visits.
All statistical analyses were performed by the R 3.3.1. statistical software (R Core Team, 2017).
A binomial generalized linear mixed model was fit to study the occurrence of metabolic disorders (dichotomized acid load and subclinical ketosis variables). The explanatory variables were ‘feed sorting status’ and production group. Production group nested in herd (group/herd) was included as a random factor, adding a random term to the intercept, that takes the correlation of data of animals from the same group and groups in the same herd into consideration.
BHB (logarithmized), NABE values and milk yield data, respectively, were compared using linear mixed models with the mentioned fixed and random effects.
Average mass proportions of TMR fractions were compared between farms where feed sorting was present and not present, using two-way analysis of variance, with time of sampling (fresh or orts), presence or absence of feed sorting, and their interaction component as fixed effects. The threshold for significance was set at P < 0.05.
Results
TMR separation
The pre- and post-feeding proportion of middle and lower sieve fractions did not differ significantly in any of the studied samples. The proportion of the smallest particle size fraction in orts was considerably reduced while the proportion of the longest particles increased in all sampled groups on 12 farms (Table 1). Based on the criteria of Heinrichs (Reference Heinrichs2013), all production groups on 12 farms were identified as feed sorting groups, and all production groups from the other 10 farms were identified as non-sorting groups. There were no farms where feed sorting status differed between the studied production groups.
Table 1. Proportion of TMR fractions on different levels of the separator (%, mean ± sd)
Metabolic disorders
Feed sorting was associated with an increase in the risk of ketosis, as compared to non-sorting. Odds ratio (OR) of ketosis was on average 5.6 times higher among fresh cows (95% CI: 1.3–24.8; P = 0.0154), 3.8 times higher in cows in early lactation (95% CI: 1.2-11.6; P = 0.0145) and on average 3.5 times higher in cows in peak lactation (95% CI: 1.04–12; P = 0.0393) (Table 2).
Table 2. BHB and NABE concentrations, prevalence of subclinical ketosis and acid load according to feed sorting and days in milk
BHB concentrations in fresh and early lactating cows were an estimated 0.6 and 0.2 mmol/l higher, if feed sorting was present, as compared to non-selecting ones (P < 0.01). Due to non-normality of data, summaries are presented as median and range (Table 2).
The odds of subclinical acidosis in association with feed sorting were on average 6.7 times higher among fresh cows (95% CI: 1.3–33.2; P = 0.013), more than 3 times higher among early lactating cows (OR: 3.3, 95% CI: 1.0–11.1; P = 0.047) and nearly 6 times higher among cows in peak lactation (OR: 5.9, 95% CI: 1.7–21; P = 0.002), as compared to cows from non-sorting groups (Table 2).
Average NABE concentrations did not significantly differ in association with feed sorting in any of the groups, except fresh cows. It was 73 mmol/l lower than in the group of fresh cows where feed sorting was present, compared to non-sorting (95% CI: 15–130 mmol/l; P = 0.0063) (Table 2).
Average milk yield of early-lactating and high-yielding cows from feed sorting vs. non-sorting groups was numerically, but not significantly different (P = 0.19 and P = 0.34, respectively) (Table 2).
Discussion
If the preference of smaller size particles over longer chop length forages causes sorting of the TMR, the nutritive value of the consumed dry matter is suboptimal (Leonardi and Armentano, Reference Leonardi and Armentano2007). In this case the amount of consumed peNDF is reduced, which lowers chewing activity and increases the amount of readily fermentable carbohydrates in the rumen (Miller-Cushon and DeVries, Reference Miller-Cushon and DeVries2017), leading to increased risk of subacute rumen acidosis. As a consequence of feed sorting the nutritive value of orts differs from that of freshly distributed TMR (DeVries et al., Reference DeVries, von Keyserlingk and Beauchemin2005). Lower-rank animals (e.g. primiparous cows, animals with health issues, e.g. lameness) that eat after dominant animals have access only to sorted TMR of suboptimal composition. Higher NDF and lower energy content of orts reduces dry matter intake and not only hinders milk production (Miller-Cushon and DeVries, Reference Miller-Cushon and DeVries2017), but is also a disposing factor for ketosis (Bjerre-Harpøth et al., Reference Bjerre-Harpøth, Friggens, Thorup, Larsen, Damgaard, Ingvartsen and Moyes2012). In association with feed sorting, we observed a higher prevalence of elevated plasma BHB concentration (subclinical ketosis) and lower urinal NABE (subacute acidosis, Humer et al. Reference Humer, Aschenbach, Neubauer, Kröger, Khiaosa-ard, Baumgartner and Zebeli2018) in fresh and early lactating cows. The increased incidence of metabolic disorders suggests imbalanced rumen fermentation due to either suboptimal or excess concentrate intake which could be linked to the presence of feed sorting. Imbalanced energy and nutrient intake can also affect milk production (Miller-Cushon and DeVries, Reference Miller-Cushon and DeVries2017). In a study on production efficiency in relation with feed sorting it was found that every 1% increase in the selective consumption of fine particles (<1.8 mm) entails a 3% decrease in milk production efficiency (produced milk/dry matter intake). The acidifying effect of smaller grain particles on the rumen fluid reduces nutrient utilization efficiency (Sova et al., Reference Sova, LeBlanc, McBride and DeVries2014). In the present study, differences in milk yield could not be detected, presumably due to lack of milk yield data of fresh and early lactation cows.
We observed that feed sorting was present either in all or none of the studied groups in a given farm, that suggests that causes of feed sorting mainly derive from feeding technology. Scientific evidence (reviewed by Miller-Cushon and DeVries, Reference Miller-Cushon and DeVries2017) and practical experience may help implementing a correct TMR and feed bunk management. Increasing amount of forages, especially the long and dry particles, low moisture content and prolonged feeding intervals promote sorting, whereas ensuring proper chop length and adequate mixing time in the diet feeder leads to a more homogenous TMR and less feed sorting. Wetting to an optimal DM content of 40–50% also reduces selectivity. The frequency by which TMR is delivered daily also influences feed sorting. Providing feed more than once per day gives less opportunity for cows to manipulate and sort the TMR.
It is important to recognize some limitations. The measurements we made on TMR samples involved only the physical composition and provided no information on nutrient composition, which may also have differed between fresh and sorted feed. Our milk yield data derived from the last milk recordings, and was not available for cows of less than 30 d pp. Furthermore, the low number of primiparous cows did not make it possible to control for parity during statistical analysis.
In conclusion, TMR quality affects animal health and production. TMR composition and homogeneity are of key importance in dairy nutrition. Ketosis and subclinical acidosis are multifactorial metabolic disorders that can also occur independently from feed sorting, however, feed sorting seems to be a predisposing factor for imbalanced energy status. TMR separation should be routinely included in herd health monitoring to detect feed sorting.
Acknowledgements
The research was supported by the EU and co-financed by the European Social Fund (EFOP-3.6.1-16-2016-00024). Viktor Jurkovich was supported by the János Bolyai Research fellowship by the Hungarian Academy of Science (BO/29/16/4).
